behavioral marxism i: collective action

7/30/2019 Behavioral Marxism I: Collective Action

1/24

Behavioral Marxism I: Collective Action

Jeffrey P. Carpenter & Peter Hans MatthewsDepartment of Economics

Middlebury CollegeMiddlebury, Vermont 05753

September 20, 2003

Abstract

1 Introduction1

Conict is of course central to Marxian economics. Consistent with Marx andEngels (1848, 3) belief that the history of all hitherto existing society is thehistory of class struggles, most, but not all, of the conict described involvesclasses, either fundamental or subsumed (Resnick and Wolff 1982). Someof it involves individuals, however, on one or both sides, often in di fferent classpositions but sometimes not. That is, the capitalists and workers who animateMarxs economics can, and do, pursue both individual and collective interests.

The individual worker is sometimes in conict with other workers, sometimeswith individual capitalists and sometimes, either alone or with other workers,with capitalists as a class. There is of course nothing wrong with this per se. Aseven staunch critics (Olson 1965), for example) acknowledge, common interestsdo, from time to time, produce common action: strikes sometimes succeed, andrevolutions sometimes occur. If there is an agency problem (Matthews 2000a)in Marxian economics, it is a contextual, rather than universal, one.

Furthermore, the co-existence of individual and collective actors, assumedor otherwise, was once not unusual, even outside the Marxian tradition. Rosner(1998) reminds us, for example, that this was also a feature of the historicalschool and, on this basis, identies Marx as a German, rather than British,economist. There are also antecedents in the British classical tradition, however:Ricardians, for example, tended to rationalize the maintenance, until 1846, of

the Corn Laws as an expression of landowners collective interest. As an his-torical matter, then, Olson (1965, 102) related claim that the Marxian model

1 We thank Julia Assael and Marla Weinstein for invaluable research assistance, Sam Bowlesfor comments on the direction of this project, and the National Science Foundation for nancialsupport.

1


2/24

of the state as the institutionalization of capitalists collective interests sets itapart from the British classical tradition misses the mark.

The choices that confront Marxs capitalists and workers, either as individu-als or as members of a coherent class, are seldom Walrasian in the sense thatthe strategic environment is a datum. Even in competitive labor markets, forexample, transactions are sometimes contestable (Bowles and Gintis 1993).As a result, a substantial and diverse (Elster 1982 or Coran 1995, for example)number of radical political economists have concluded that both the conictsand the formations on either side could be represented in game theoretic terms.Even in its current, still primitive, state, for example, the literature on socialdilemmas has much to tell us about the conditions under which class rationalbehavior is, and is not, a sensible abstraction.2

Indeed, to the extent that the modern Marxist characterization of the stateas the mechanism for the collective action of capitalists can (at least some-times) be rationalized, the force of Elster (1982) criticisms of OConner (1973),Foster (1973) and others as weak functionalism is somewhat blunted. (Thesame cannot be said for (Elster 1982) non-Marxist functionalists: Posner (1977)economic interpretation of the law, for example, is an exercise in strongfunctionalism.)

The search for plausible microfoundations for collective action does not re-quire a commitment to strict methodological individualism. It will sometimesbe the case, as Kirman (1997) observes, that even perfect, but probabilistic,information about individual behavior is not sufficient to predict macrostruc-ture. The observation that some spontaneous orders are difficult, or perhapsimpossible, to rationalize in these terms does not mean that we should neverattempt to do so. Indeed, it is important for radical political economists to(re)claim the term. As Blume (1997) reminds us, there is no argument that

the spontaneous order of the Austrians is necessarily bene

cent . . . [t]he invisi-ble hand could equally well be Shakespeares bloody and invisible hand of nightas Adam Smiths hand of Pangloss.

If the strategic considerations embedded in most Marxian treatments of in-dividual and class conict seem to underscore the need for game theoretic micro-foundations, an important problem must rst be overcome: even in controlledenvironments, individuals often do not behave as standard game theoretic mod-els would predict. It is for this reason that Muntaner (1994, 118-9) and othershave dismissed rational choice Marxism, game theoretic or otherwise, as a for-malistic and deductive [enterprise] . . . often contradicted by . . . [the] data.From this perspective, the existence of class as a primitive of sorts becomes analternative to individualist models. Following Burawoy (1989), Muntaner (1994,118) argues that it is the absence of a realist theory of knowledge that under-

mines rational choice Marxism, in contrast to other behavioral traditions in thesocial sciences (e.g. social psychology), whose factual and inductive characteris the result of its commitment to experimentation.

2 Social dilemma is a broad term for any interaction in which group incentives di ffer fromindividual incentives. Examples include collective action, the provision of public goods, andthe maintenance of commonly held property.

2


3/24

We reach a different conclusion, however. In particular, we believe that therecent literature in experimental economics (together with previous behavioral

work outside economics which economist now take seriously) has moved thediscipline (much) closer to Muntaner (1994) ideal, and that observed behaviorin the lab is (much) closer to Marxs conception of economic man than homoeconomicus. This does not mean, of course, that the individuals involved arenot purposeful, as opposed to hard nosed (Ledyard 1995) rational, or oftenself-interested. Furthermore, we believe that it is, or soon will be, p ossibleto rationalize class rational behavior in these terms, and that much of thisbehavior is consistent with evolutionary game theory. We call this researchproject behavioral Marxism.

2 Class Consciousness and Collective Action

Consider Elsters (1982; 1985) canonical example of Marxian functionalism, itsrationalization of the observed bias in technological change. As he representsthe argument, capitalists as a class benet from the search for, and introductionof, labor-saving methods of production to the extent that these exert downwardpressure on real wages. It does not therefore follow, however, that individualcapitalists have an incentive to do so, because no one capitalist can inuence thewage rate on his/her own, and because each has an incentive to free ride onthe efforts of others to do so. The bias, he determines, cannot be explained interms of its benets to capitalists as a whole, but the conclusion is not denitive:if it exists, class consciousness [must be] the capacity of a class to behave asa collective actor . . . to overcome the free-rider problem Elster (1982, 466).This is more or less consistent with Mansbridge (forthcoming) denition of anoppositional consciousness, itself intended to echo Marxs (1852) notion of a

class for itself (fr sich) rather than in itself (an sich).Almost two decades earlier, Olson (1965, 105) had been unequivocal: the

absence of the sort of class action Marx predicted is due in part to the pre-dominance of rational utilitarian behavior, [f]or class-oriented action will notoccur if the individuals that make up a class act rationally. One reason for thedifference is that Olson (1965) frames the collective action problem as a oneshot prisoners dilemma, in which each worker (or capitalist) has a dominantstrategy: if other workers engage in class rational behavior, then thosewho abstain can free ride, but if others abstain, then abstention is alsopreferable to unilateral engagement. In contrast, Elster (1982; 1985), citing Sen(1973) interpretation of the Critique of the Gotha Programme, understood thatthe collective action problem is sometimes better represented as an assurancegame (Runge (1984) with multiple Nash equilibria. Others have formalizedit as a discrete public good game (Diermeier and Van Mieghem 2000) or as apublic good game with individual thresholds (Granovetter 1978), all of whichalso exhibit multiple equilibria.

Even as a one shot prisoners dilemma, however, the collective action prob-lem is more complicated than rst seems. First, as summarized in (Ledyard

3


4/24

1995), there is now a wealth of data on public goods games in which a non-negligible fraction contribute in the rst and nal round(s) of an experiment,

despite hard nosed predictions. Second, and more important, one shot revo-lutions sometimes do occur. Oliver (1993, 273) concludes that if he had beena sociologist, Olson might [then] have . . . launch[ed] a theory of the nonrationalor nonindividualist bases of collective action. At the least, if one denes ra-tional in hard-nosed terms, then individualist models of collective action insimple games should impose bounds on this rationality.

With or without such bounds, the dismal logic of the prisoners dilemma isundermined when the collective action problem is repeated. Friedman (1971)and Taylor (1976) were perhaps the rst to rationalize what Hardin (1982) andothers observed in practice, that abstention is not a dominant behavior whenthe discount rate is small enough, or the future matters enough, to supportconditional co-operation. Indeed, this Folk Theorem has since been extendedto cases in which public information about the behavior of others is imperfect((Fudenberg, Levine et al. 1994)).

It is this framework that also allows the rst part of Sabia (1988) defenseof the Marxian model of collective action to be formalized, even for homo eco-nomicus. Drawing on the characterization of the French peasant class in TheEighteenth Brumaire (Marx 1852), Sabia (1988, 56) identies the existenceof multiple, stable and ongoing relationships between . . . like-minded workerswithin . . . small groups as a precondition for the emergence of class conscious-ness. As he earlier describes this stage of development, Marxs view is thatconcentration, homogenization, and enlightenment engenders not only a growingconsensus on the part of ever-growing number of workers about their situationsand needs but propels also the creation of local worker associations and organi-zations within proletarian communities (Sabia 1988, 54). Without the concen-

tration of workers into towns and factories, for example, the manifold relations(Marx 1852, 317) that are a prerequisite for conditional cooperation in repeatedgames cannot exist, and it was for this reason that the French peasants remaineda class in, but not for, itself. In a similar vein, the homogenization of workerswould eliminate, or at least mitigate, the problem of imperfect information, an-other obstacle to engagement. Last, stripped of its normative connotations, theenlightenment of workers amounts to a requirement that workers understandthe benets (and costs) of collective action.

It is important to note, however, that the engaged outcome is one in a con-tinuum of sustainable equilibria for the repeated collective action game. Inparticular, abstention in each period remains an equilibrium. In more formalterms, absent an equilibrium selection mechanism, the mobilization of even smallgroups of workers (or capitalists) is not inevitable. This is perhaps as it should

be. In all but the most mechanistic interpretation of Marxs work, concentra-tion, homogenization and enlightenment are not sufficient conditions for thedevelopment of class consciousness.

But how does the expression of reciprocal behavior in small(ish) groups,when it exists, evolve into a commitment to much larger classes? Bendorand Mookherjee (1987) nd, for example, that even with perfect information

4


5/24

about the total contribution to the collective good, there is, for each discountrate, the number of participants for which conditional co-operation is possible is

bounded above. (This result is sensitive, however, to the production functionfor the collective good.) With less information, or with unobservable differencesin individual costs of contribution, the problem becomes even more acute.

Sabias (1988) concludes that the reciprocal but self-interested strategiescharacteristic of small clusters, rst established in the icy water[s] of egotisticalcalculation, can sometimes become the conventions or norms that can supportcollective action in much larger groups. That is, the behavior that was oncesufficient to cause homo economicus to punish a free rider can become a normabout, for example, injustice, and the basis (Mansbridge 2000) for oppositionalconsciousness. Homo economicus is transformed into homo reciprocans (Bowlesand Gintis 1998).

It remains to show, however, that these conventions, once established, cansurvive in a world where individuals will sometimes mutate into free riders,or in which these conventions are sometimes put to the test. From time to time,norms must be enforced, and enforcement requires sanctions. Sabia (1988, 57)believes that even as local associations expand, or are somehow connected toone another, that effective sanctions are possible because any potential freerider would be violating a convention that he or she respects at the smaller leveland because some others in the smaller group(s) of which [he or she] is partwill know this. On the other hand, Oliver (1993) and others are suspicious ofsolutions to the collective action problem that require, or seem to require, theprovision of a second order collective good. That is, even norm adherents willhave an incentive to free ride on its enforcement.

Recent developments in evolutionary game theory suggest this could be asmaller problem than rst seems, however. Consider a two stage game in which

the

rst stage is the standard prisoners dilemma. In the second stage, the

rststage choices are revealed and participants are able to punish one another, atsome cost to themselves. In this framework, then, a co-operative norm enforceris someone who engages in collective action in the rst stage and sanctionsthose who abstained. Combined, the two stages constitute the norms gamerst described in Axelrod (1986). Suppose that participants in this norms gameare boundedly rational in the sense that each is committed, perhaps for thereasons Sabia (1988) describes and we elaborate on in section 4, to one of theeight possible pure strategies, and that, for the moment, participants are re-paired, at random, so that the conditional reciprocation consistent with thevarious Folk Theorems is ruled out a priori. Last, the participants are assumedto be self-interested in a purposeful, rather than hard-nosed rational, sense:each has an aspiration level drawn from some uniform pdf, and does not

alter her/his behavior if her/his payoff exceeds this aspiration, but switchesto another, with the same likelihood as its current share in the population,otherwise. Under these conditions, the so-called replicator dynamics (Taylorand Jonker 1978) will describe the evolution of norms.

It is not difficult to see that, under these conditions, a population of normenforcers will be neutrally stable. So, too, will a population of unconditional

5


6/24

defectors, so that the formation of class consciousness is possible, but not in-evitable, and will turn on the initial shares hence the importance of Sabias

(1988) local associations and the respective basins of attraction.Sethi (1996) adds a ninth strain, homo economicus, who chooses the hard

headed best response to each possible opponent, with surprising results. First,and most important, there are no evolutionarily stable states (ESS), or forthat matter neutrally stable states (NSS), in which homo economicus surviveson her/his own. If the hard nosed survive, it is in an ESS in which eitherbullies those who abstain from collective action and punish those who donot or passive defectors those whose commitment to the rational choice(abstain, and refrain from punishment) is unconditional coexist with them.Radical political economists will not nd it difficult to visualize either scenario.In the rst, enlightened self-interest characterizes some workers (or capitalists)who would be prepared to contribute to collective action if a su fficient numberof others were committed, in a non-rational sense, to the cause, but theseworkers never mobilize because of the presence of anti-collectivists, who donot contribute and punish those who do. In the second, the same enlightenedworkers or capitalists do not mobilize because there are a substantial numberwho understand self interest in much cruder terms.

Second, there is an ESS in which co-operative norm enforcers alone survive.Furthermore, participants fare better in this environment than either of theother two. That is, there are environments in which class consciousness canexist, even thrive, and the process that produces this outcome has plausible,individualist, microfoundations. Whether or not it happens depends, once more,on historical preconditions initial values - and chance events.

Furthermore, if and when this class conscious equilibrium is reached, smallperturbations in the strategic environment will lead to behavior consistent with

the collective pursuit of common interests. In Matthews (2000b), the samenorms game is recontextualized as a variant of the Michl and Baldani (2000)model of technical change as a prisoners dilemma, to show that even undervariations in cost conditions, the choices of class conscious capitalists will some-times be consistent with their collective interest, in this case the maintenanceof the prot rate in the face of Marxs tendential law.

Carpenter and Matthews (2002) extend Sethis (1996) model to allow forboth in group and out group sanctions, and nd that social reciprocatorscan also survive.

There is a second, albeit related, set of models that could allow class con-scious behavior to be rationalized. In indirect evolutionary models, it is pref-erences, rather than behavioral rules, than unfold over time. Fitness is stillmeasured in terms of material benets, but this now determines the rates at

which preferences, rather than behaviors, are transmitted. These models seemcloser in spirit to Sabias (1988, 59) characterization of class consciousness, inwhich solidarity [should be] understood as a form of consciousness or compo-nent of individual character rather than a selective incentive but, followingBinmore, McCarthy et al. (2000), Sethi and Somanathan (2000) conclude thatthere are methodological difficulties with this approach.

6


7/24

Viewed from this perspective, the question becomes if, and how, a preferencefor class-based solidarity might survive under a plausible selection mechanism.

The answer, once more, seems to be that there are (some) equilibria in whichthis occurs. In Guttman (2000), for example, participants are able to drawinferences about, and exploit, the preferences of others. Even if preferencescannot be observed, however, assortative interaction is sometimes possible, asin Bowles and Gintis (1999).

3 A (Very) Simple Model

Suppose that there are N workers, each of whom must choose how much of theirendowment w to contribute to the pursuit of class-related objectives. For eachcontribution xi, all receive rxi, where r < 1, so that the material payoff to i is

just i = wxi + rP

xi. If, in addition, rN < 1, this collective action game is

a prisoners dilemma, in the manner of Olson ( 1965). Standard game theoreticmethods predict (in the one shot version of the game, at least) that no workerwill contribute.

Suppose, however, that there is some, perhaps small, likelihood of a shockto workers preference functions. That is, it is now possible that workers, per-haps as a result of their participation in local associations, have assimilated areciprocal norm or convention, but that it is also possible the anonymous na-ture of class membership causes workers to overestimate the benets of selshbehavior. As Anderson, Goeree et al. (1998) then show, if the structure of theseshocks assumes the continuous version of the logit (Chen, Freidman et al. 1995)form, then each workers prior pdf over actions will be:

f(xi) =exp(e

i(xi)/)

Rexp(ei (xi)/)dx

where is proportional to the standard deviation of the error distribution.If each worker is rational in the sense that s/he assumes other workers will

experience similar shocks, then the equilibrium is a xed point of the functionthat maps distribution of actions into expected utilities and expected utilitiesinto distributions of actions. Anderson et al. (1998) show that in linear publicgood games like this one, the quantal response equilibrium is:

f(x) = exp()

1 exp(w)

where is proportion to 1 r. The mean contribution is between 0 andw/2, and the equilibrium has four important properties. First, an increase in

r, the value of the public good, leads to an increase (in stochastic terms) incontributions. Second, an increase in the size of the preference shocks also leadsto an increase in contributions. Third, an increase in the number of workers hasno effect on individual contributions, and fourth, an increase in endowments wis associated with an increase in contributions.

7


8/24

All four properties have important implications for the evolution of classconsciousness. The rst suggests that even self-interested workers whose prefer-

ences are vulnerable to small trembles as a result of their involvement in localassociations will contribute more as the benets of mobilization increase. Thelikelihood of rational revolution rises, in other words, when the differencesbetween capitalists and workers are wide. In a similar vein, the more successfullocal associations have been, the larger the standard deviation of the preferenceshocks, and the more workers will contribute to broader causes. The third re-veals the Olsonian conjecture about the effects of group size to be more fragilethan often believed: mean contributions do not fall as the number of workers Nrises. And last, workers who can contribute more often will, other things beingequal. The rst and last hint that the timing of collective action is subject tocompeting pressures: in the earliest stages of capitalism, when the differencesbetween capitalists and workers are wide, workers also have less (in absoluteterms) to contribute.

Under the usual interpretation, is a monotonic measure of the likelihoodthat workers make errors: as tends to zero, no one contributes, and as ittends to innity, workers pdf over actions becomes uniform. It is our position,however, that what lies between homo economicus and random choice is notnear rational behavior but something else, an awareness of the possibilitiesfor collective action (we term this an ingroup bias).

If one substitutes a step function or provision point for the productionof the public good that is, for workers to revolt, the number who engagein collective action must exceed some threshold the results are even moredramatic. Using the same sort of perturbations as Anderson, Goeree et al.(1998), Diermeier and Van Mieghem (2000) nd that the model predicts rare,sudden bursts of collective action that tend to occur in clusters, consistent,

in their view, with the1

989 Leipzig Monday demonstrations that led to thecollapse of the GDR.Furthermore, McBride (2001) solves for the quantal response equilibrium of

a model where the threshold is unknown to participants, and nds that undersome conditions, wider uncertainty can even be desirable.

Chwe (1999), on the other hand, allows individuals to be hard-noised, butassumes that each has her/his own participation threshold, a formalization ofGranovetter (1978), and embeds each within a social network. He nds thatwhen thresholds are low, strong links, of the sort forged among friends in localassociations, facilitate participation, but when thresholds are high, weak linksare preferable. Furthermore, he concludes that the conventional wisdom thatcollective action is sensitive to the thresholds of the rst, or earliest, individualsto move assumes the absence of reciprocal information.

8


9/24

4 Empirical Support for a Marxian Theory of

Collective ActionIn this struggle, of which we have noted only a few phases, this massbecomes united, and constitutes itself as a class for itself. The inter-ests it defends become class interests.(Marx, 1977, Selected Writings, p.214)

Although a small literature has arisen that can be described as Marxiansocial psychology (e.g. Leontev 1968, Young 1975, and Ulman 1991) that ad-dresses exploitation and inequality in addition to collective action, we follow adifferent path seeking specic behavioral evidence to justify our claims aboutthe plausibility of a Marxian theory of collective action from all the social sci-ences. Specically, in this section we discuss experimental evidence from social

psychology, sociology, political science, and economics. Further, our analysisfocuses on six factors which Marx asserted either favored or hindered collectiveand class action. The six factors we concentrate on expand on the list discussedabove and are: the existence of an ingroup bias, actors understanding of so-cial dilemmas, group homogeneity, group turnover, the structural componentsdetermining the productivity of prosocial acts, and social sanctions.3

4.1 Ingroup Bias

Marx asserted the differential propensity of class members to behave prosociallytowards each other when common interests where are stake. Workers, for exam-ple, feel solidarity with other workers and this causes them to be more likely tocooperate in face of the incentive to free-ride. Sociologists and social psychol-

ogists refer to this phenomenon as the existence of an ingroup bias (IGB). Ingeneral, an IGB occurs when members of a well-dened group behave differentlytowards other members of the group than they do towards people outside thegroup. For our purposes, IGB is important because it manifests itself in socialdilemma situations by causing group members to be more cooperative with eachother.

The behavioral evidence supporting the existence of an IGB is expansive.The most robust nding is that forming or increasing the salience of an ingroupleads to more within-group cooperation in prisoners dilemma, public goods,and common pool resource experiments. In her survey of ingroup experiments,Brewer (1979) concludes that becoming a member of a group causes a person tolook favorably on other group members and care more for their well-being. Fur-ther, she reports that this prosocial effect is a stronger determinant of behavior

than the associated process by which group members simultaneously decreasetheir opinion of outgroup members.4

3 A fuller development of these factors is presented in Elster ( 1985) and Sabia (1988).4 Also see Eckel & Grossman (2001) who generate an IGB in a team production setting

and Yamagishi & Kiyonni (20000) who show players in a prisoners dilemma situation expectmore trust and reciprocity from ingroup members.

9


10/24

The existence of an IGB also appears to interact with other variables thatMarx saw as determinants of the propensity toward prosocial acts. For example,

Brewer & Kramer (1986) show that group size, a causal factor of the productivityof prosocial acts, interacts with the salience of group boundaries in a commonpool resource experiment. Counter to the conventional logic, larger groupsshowed more restraint when group affiliation was stressed. Similarly, Brewer(1979) writes that more homogeneous groups (where homogeneity is measuredby ideology as well as other factors) are more likely to demonstrate and actin accordance with an IGB. Finally, Kollock (1998) argues that an IGB cancloud ones understanding of the incentives of a social dilemma situation. 5 Inhis prisoners dilemma experiments, members of highly salient ingroups rankedmutual cooperation over defecting on a cooperator despite the greater materialgain to defecting. At the same time, as the salience of group membership fell,more players reversed this ranking.

In addition to identifying the existence of an IGB, experiments have alsobeen conducted to identify the microfoundations of the bias.6 One dimensionof the IGB that stands out is conformity. For example, Parks, Sanna et al.(2001) demonstrate that participants in a hypothetical social dilemma tend toconform to the behavior of the other members of their ingroup. Further theyshow that this behavior arises in both large group public goods games and two-person prisoners dilemma games. In a different setting, Haslam and Platon(2001) show that, to be effective at organizing collective action, leaders mustconform to and reinforce conformity to group ideology. It appears that increasedconformism translates into a more salient IGB and more group-related prosocialactivity.7

Not only does an IGB lead to more prosocial activity within the group,as stressed by Marx in the context of workers and capitalists, an IGB also

leads to feelings of aggression and competition towards outgroups. Komorita& Lapworht (1982), Kramer & Brewer (1984), Bornstein et al. (1996), andCarpenter and Cardenas (2001) all show that splitting a group that faces a socialdilemma into subgroups causes competition to arise between the subgroups tothe detriment on the overall level of cooperation. Similarly, Mackie, Devos et al.(2000) report that the stronger group members feel the group position is (i.e.in this case the more salient the ingroup is),the more likely they are to supportaction against an outgroup.

5 Ahn, Ostrom and Walker (1999) nd similar results without checking for an IGB.6 Note, in this case, by microfoundations we mean behavioral attributes of individuals that

lead to ingroup biases. We dont assume these attributes are the result of a decision makingprocess in which cost and benets are weighed. That is, the current meaning differs fromstandard usage - providing incentive compatible explanations of macrophenomena.

7 Prosocial acts within groups are not limited to cooperation in social dilemmas. Carpenter,

Burks and Verhoogen (2001) show that workers demonstrate considerably more altruism to-wards each other than students do towards each other in a Dictator experiment. When given$100, students give away $25 on average while workers give away $45. One can argue thatthe salience of the workers group membership is higher because of repeated daily interactionsand the relative importance of cooperation in the interactions on top of the mostly social tiesstudents tend to form.

10


11/24

0

1

2

3

4

56

7

NoGroup

Frate

rnity

Greek

UCLA US

CPo

lice

(C,C)

(D,C)

(D,D)

(C,D)

Figure 1: Naturally Occurring IGB (source: Kollock 1998)

As it is representative of this research, we believe it is useful to explainKollocks experiment in more depth. Participants were told that with someprobability they would play a prisoners dilemma game with someone else afterlling out a survey in which they ranked the four outcomes of the games normalform: (C,C), (D,C), (D,D), and (C,D). To manipulate the degree of IGB, theparticipants (fraternity members at UCLA) were told they were playing with ananonymous other from one of four groups: fellow fraternity brothers, studentsfrom different fraternities and sororities on campus, students from a rival campus(USC), or the UCLA Police Department.

Figure 1 summarizes Kollocks results and provides evidence that naturally

occurring group boundaries make group membership salient and trigger an IGB(and and outgroup bias). The main, signicant, effect of group membership wasin the ranking of the mutual cooperation outcome, (C,C), (p


12/24

the group. This bias takes the form of changes in individual social orientation(Griesinger & Livingston 1973, McClintoch & Liebrand 1988) in which one be-

comes more altruistic towards others in the group and more competitive towardsindividuals not in the group. These biased preferences then, under the right cir-cumstances, translate into more cooperation within groups and more aggressiontowards other groups.

As we will see, the other factors we attribute to Marx can be, to one degreeor another, subsumed under the ingroup bias. That is, with further review ofthe behavioral literature, it will become clear that homogeneity (as we have seenabove), understanding the incentives of a social dilemma, turnover, punishment,and structural factors all can be seen both as determining the propensity tocooperate directly and as affecting the formation and prominence of an ingroup.This means that our re-interpretation of a Marxian theory of collective actionhypothesizes that one factor is crucially important, the IGB. As in section 3above, we claim that strengthening a classes IGB translates into the kind ofsolidaristic preferences necessary for our quantal response model to be plausibleand select equilibria in which class action obtains. Additionally, while the focusis on the IGB, we complicate the analysis be noticing that the other ve factorsmatter both to the degree that they make group boundaries salient and throughother, more direct, channels.

4.2 Understanding Ones Situation as a Social Dilemma

They [the working class] ought to understand that, with all the mis-eries it imposes upon them, the present system simultaneously en-genders the material conditions and the social forms necessary foran economic reconstruction of society.

(Marx,1

975, Wages and Pro

ts, p.1

52)There are (at least) two ways in which workers might misunderstand the

incentive to act collectively. First, they may simply lack the cognitive ability tounderstand their class position (i.e. that there are potential benets to prosocialacts, but no one has the unilateral incentive to act prosocially) and it may notbe clear where to direct class action (e.g. workers may not know that capitalistsexploit them). Presumably, this cognitive problem will diminish the likelihoodof class action, but can be corrected by education. However, misunderstandingthe logic of collective action might also work in favor of the working class becausethe source of misunderstanding, framing and an IGB, tends to make people morecooperative.

Indeed, ones cognitive capabilities have been shown to affect play in the

prisoners dilemma, but the results are counter-intuitive (at least to economists).Nydegger (1974) and Pincus and Bixenstine (1979) show that people who arebetter at abstract information-processing are more likely to cooperate in theprisoners dilemma. In general, these studies show that greater cognitive abilitytranslates into a higher propensity to cooperate. However, this result is curiousbecause one would imagine that higher cognitive ability would make one better

12


13/24

at game theory, and if people are basically egoistic, better game theorists aremore likely to defect in a nitely repeated or one-shot prisoners dilemma.

Framing may affect both ones cognitive ability to understand social dilem-mas because different frames either hide or accentuate different relationshipsand ones processing of the incentives of collective action because frames maymake group boundaries either more or less salient. Pruitt (1967; 1970) andPincus and Bixenstine (1977) look at the decomposed prisoners dilemma to un-derstand the effect of framing. Table 1 illustrates an example of a decomposedgame.

Matrix A Matrix B Matrix CGive Give Give Give

A B me him me himA 12,12 0,18 A 6 6 A 0 12

B1

8,0 6,6 B1

2 -6 B 6 0Table 1: The Decomposed Prisoners Dilemma (source: Pruit 1967)

Matrix A is the standard representation of the game where strategy A iscooperative, but dominated by B. Matrix B is one decomposition of matrix A.Notice, if both players choose A, they both keep 6 and give the other person 6yielding the (12,12) outcome in matrix A, and if one chooses A and the otherB, the B chooser takes 6 from the A chooser and adds it to his or her outcomeresulting in the (18,0) or (0,18) outcome. Matrix B is thought to emphasizethe control one has over gains while Matrix C is thought to emphasize theimportance of mutual cooperation. When played, matrix A, the fully composedgame elicits 40% cooperation while the individualist game, matrix B, elicits lowlevels of cooperation (20%) and the collectivist game, matrix C, generates high

levels of cooperation (80%).One interpretation of the data on the decomposed prisoners dilemma is

that some how matrix B is a better description of the incentives to a layper-son because it elicits less cooperation. Alternatively, one could argue that thedifferences in the frames either make the two players feel more or less like agroup and from section 4(a) we know that group members are more likely to actprosocially towards each other. In this case, presenting the game in a way thatemphasizes the need for cooperation to achieve a high payoff (matrix C) maymake players feel as if they have common cause and are members of a commongroup while emphasizing the ability to get 12 unilaterally by choosing B (matrixB) may hinder solidaristic feelings among participants.

4.3 Group Homogeneity

The ordinary English worker hates the Irish worker as a competitorwho lowers his standard of life. . . The Irishman pays him back.. . Hesees in the English worker at once the accomplice and the stupid toolof the English rule in Ireland. This antagonism is articially keptalive and intensied by the press, the pulpit, the comic papers, in

13


14/24

short, by all the means at the disposal of the ruling classes. Thisantagonism is the secret of the impotence of the English working

class, despite its organization.(Marx, 1870, to Meyer and Vogt)

Group homogeneity is linked to the ingroup bias for obvious reasons het-erogeneous groups tend to form into subgroups that, as we saw in section 3(a)compete with each other. According to Brewer (1979), in general, the more ho-mogeneous a groups is, the more salient is the IGB and the more group benecialis the activity that occurs.

To examine the effect of an ingroup bias and outgroup competition, Carpen-ter & Cardenas (2001) examine how differences in group affiliation may affectthe level of cooperation in commons situations. To do so the authors designa real-time, cross-cultural common pool resource (CPR) experiment purposelyusing participants from cultures that derive different benets from biodiver-

sity (extraction versus conservation) to analyze the effect of group affiliationon cooperative behavior. In the CPR environment, the authors nd evidencethat group affiliation affects behavior. Specically, they show that Americanstudents maintain their extraction in the mixed treatment (both Colombianand American participants) compared to homogeneous groups (American only)while Colombian participants extract more in the mixed treatment.

However, the authors also witness an outgroup bias which takes the form ofcompetition and negative reciprocity by exploited subgroups. Here subgroupsthat extract less in one period (i.e. are exploited) tend to extract more in thefuture and the magnitude of this adjustment is determined by participant na-tionality and our treatments. Figure 2 illustrates Carpenter & Cardenas mainexploitation results. Figure 2 plots the size of the regression coefficient on thevariable that measures how much a subgroup was exploited last period (i.e. howbig the difference was between the subgroups extraction level and the level ofthe other subgroup). Further, the data is split between what the authors termpositive reciprocity, when a subgroup reduces its extraction after the other sub-group does so and negative reciprocity which occurs when subgroups react spite-fully to greater exploitation (i.e. react to increased other subgroup extractionby extracting more oneself). As one can see, there is not much positive reci-procity between subgroups; in fact, none of the positive reciprocity coe fficientsis signicantly different from zero. However, there is considerable competitionand negative reciprocity. And, most interestingly only for Columbians does thelevel of competition between groups depend on whether the larger group is ho-mogeneous or heterogeneous (compare the negative reciprocity coefficients forBogota and Bogota Mixed).

However, homogeneity may also play an independent role in determining theamount of cooperation in a group. First, Stallings (1973) reports that groupaction tends to evolve from homogeneous initial beliefs according to Smelser(1962) generalized belief hypothesis in which groups are more likely to organizecollective action when they share beliefs. However, he also illustrates how theenvironmental movement has organized itself despite a signicant level of belief

14


15/24

Positive Reciprocity Negative Reciprocity

-5

0

5

10

15

20

25

30

Bogota BogMixed

Middlebury MiddMixed

Figure 2: Reactions to Exploitation (source: Carpenter & Cardenas 2001)

heterogeneity due to an endogenous group processes similar to our model ofsection 3. Second, according to the triangle hypothesis (Kelly and Stahelski1970) competitively predisposed individuals expect others to be homogeneouslycompetitive while cooperative types tend to believe the population is mixedbetween competitors and cooperators. Returning to gure 1, we see that, acrossconditions, being the sucker in a social dilemma is ranked as the worst possibleoutcome. Now, if players make expectations about the cooperativeness of thepopulation they play social dilemma games in, then the initial cooperativeness

caused by an IGB will spread because of the expectations of cooperators.There is evidence supporting the triangle hypothesis. Miller and Holmes

(1975) compare the expectations of people playing in two different dilemmas andnd that competitors do expect to be playing with a homogeneous population ofcompetitors. Further, van Lange (1992) replicates the general results of Miller& Holmes and also shows that competitors are more sure of their expectationsthan cooperators are.

4.4 Group Turnover

As with understanding the situation, we can think of at least two ways in whichgroup turnover might affect cooperation. The rst is derivative of repeatedgames. Specically, given the folk theorem, games with uncertain endpoints

and stable group membership allow the equilibria to arise in which cooperationis sustained by reputation or reciprocity and punishment (e.g. playing tit-for-tator other trigger strategies). However, such cooperative outcomes do not arisewhen groups are constantly being reshuffled because there is no incentive toform a reputation or punish free riders when you will not likely be in the samegroup later to recoup the costs of punishment.

15


16/24

To test the relevance of these sorts of folk theorem explanations of coop-eration, economists have run a number of public goods experiments in which

players either stay in the same group for the entire experiment or are randomlyre-shuffled into new groups after each decision-making round. Intuition says thatgroups of partners (i.e. those who stay in the same group) will cooperate morebecause they can form reputations and credible punish free riders and groupsof strangers (those who are re-shuffled) will cooperate less. Initially, Andreoni(1988) produced exactly the opposite result, strangers cooperated more thanpartners. However, upon replication, Croson (1996) and Keser & van Winden(2000) nd that partners do contribute more to a public good than strangers.Hence, overall it appear that group turnover may matter to the extent that itreduces the incentive to police noncooperation and invest in reputations.

However, another set of economic experiments shows how turnover mightactually benet cooperation. The key, in this case, is that if turnover is non-random and, in particular, leads to assortative interactions (i.e. cooperatorsare more likely to meet cooperators and defectors are more likely to meet de-fectors) then cooperation can ourish.8 In a clever experiment, Ehrhart andKeser (2000) allow group membership to evolve endogenously in a public goodsgame. Specically, they allow players to leave groups at some cost and formnew groups. Their results show that this manipulation does allow cooperatorsto leave groups lled with free riders and establish new cooperative colonies, butbecause cooperators can not act parochially to keep free riders out, free riderssoon nd cooperative groups and invade them. With more power to controlgroup membership, one might expect that cooperation would ourish in groupsestablished by those who signal their social orientation by leaving uncooperativegroups. In a related experiment, Ameden et al. (1998) also allow non-randomgrouping, but in this case, the assortation is done by the experimenters. The

instructions state the people will be sorted, but not how. Placing all the mostcooperative players in one group allows the high contributing group to achievenear Pareto optimal outcome levels.

However, we feel that the rationale closest, in spirit, to what Marx hadin mind was that it is harder to maintain cooperative norms in groups thatare constantly having members leaving and new members (who presumablyhave not internalized the groups norms) entering. As far as we can tell thereis little in the way of experimental evidence on one side or the other of thishypothesis. However, one experiment run by Schopler, Insko et al. (1994) isvery close in design to this concept. In this experiment, 3 people form a groupand decide collectively on a group strategy in the prisoners dilemma (theyplay against another 3 person group). The experiment is repeated and playedfor four trial blocks of ve rounds each. In the rst trial block, one of the

three group members is a confederate of the experimenters. In one conditionthe confederate is a strong advocate of cooperation, in a second condition theconfederate is passive (i.e. neither advocating nor opposing cooperation). For

8 This argument also applies from a theoretical perspective if one allows a specic form of thereplicator dynamic that allows for assortative interactions. See Skyrms (1996) or Carpenter(1999) for examples.

16


17/24

0

0.2

0.4

0.6

0.8

1

Trial 1 Trial 2 Trial 3 Trial 4

Active

Passive

None

Individuals

Figure 3: Mean Proportion of Cooperation (source: Schopler et al.,1

994)

our purposes, the interesting manipulation is that after the rst trial block, theconfederate is replaced by another subject and the game is continued. Furtherafter the second trial block another original team member is replaced and afterthe third block, none of the original members are left.

Figure 3 illustrates the mean level of cooperation in the two treatments com-pared to the case when there are no confederates (None) and when participantsplay as individuals instead of groups. The interesting comparison is the tra-

jectories of the active and passive treatments. The initial levels of cooperationare at the extremes of the elicited values with the manipulations working in thehypothesized direction (active confederates achieve more cooperation, but pas-

sive confederates dampen cooperative spirit), however by the end of the forthtrial block, the levels of cooperation are comparable. The main result is thatturnover does weaken an initially strong sense of cooperation fomented by theactive confederate. Further, the drop in cooperation (trial 1 mean minus trial 4mean) is the largest in the active condition indicating that the effects of strongopinions decay faster than group generated opinions. In this sense, Marx didcorrectly hypothesize how turnover might affect the sustainability of collectiveaction.

4.5 Structural Factors

Ledyard (1995) surveys the effect of structural factors on the provision of apublic good. By structural factors we basically mean the factors that determine

the productivity of a prosocial act. In public goods experiments, there are twofactors, group size and the productivity of the public good which is summarizedin the marginal per capita return (MPCR) from the public good. In most publicgoods games players have two choices for the use of their token endowment. Onechoice is to keep all their endowment which returns one experimental monetaryunit per token kept and the second choice is to contribute part or all of their

17


18/24

endowment to the public good which yield the MPCR for each group memberregardless of their contribution levels. For the game to simulate the provision

of a public good, one sets MPCR1. Table 2 summarizes theeffects of changing the group size between 4 and 10 persons and changing theMPCR from 0.3 to 0.75.

Average Contribution LevelMPCR

N 4 19% 57%10 33% 59%

Table 2: Structural Change iand Public Goods (source: Ledyard 1995)

In general, increasing the MPCR increases contributions despite contribut-ing nothing still being the dominant strategy. While this fact is anomalous

theoretically, it makes sense that reducing the bene

t to free riding (1

-MPCR)should result in less free-riding as Marx would have predicted. Perhaps moreanomalously, one can also see that increasing the size of groups also increasescontributions contrary to Olsons (1965) hypothesis. However this result is par-tially contrived because, notice one cannot change the size of the group withoutimplicitly changing the relationship between the incentive of the group and theincentives of the individual. For example, say group size is four and the MPCRis 0.3. In this case, an individual contribution generates 0.34=1.2 EMUs inbenets for the group, but if we increase the group size to ten the a contri-bution generates 3 EMUs in benets. Hence, in this experiment larger groupscan take advantage of higher payoffs and specically, if only 4 of the 10 con-tribute, individuals gain 1.2 EMUs from the public good which doesnt look sobad compared to the 1 EMU one gets from keeping another token. However,

in four person groups, everyone has to contribute before the public good seemslike a good investment.

Overall, changes in the incentive to contribute to collective action do seemto matter to individuals. Despite there being a strong, Olsonian predictionthat everyone will free-ride, reducing the benet to free riding does increasecontributions.

4.6 Punishment

We are also interested in the role of punishment. Punishment opportunities,often designed by experimenters to be empty threats, have been used regularlyby participants to sanction players acting with self interest. In relation to aMarxian theory of collective action and class consciousness, cooperation can be

supported if people punish free riders.The rst public goods experiment incorporating mutual monitoring was Fehr

and Gaechter (2000)9 who conrm the reciprocity conjecture generated by An-

9 However, using the broader category of social dilemma experiments, Ostrom, Walker andGardner (1992) were the rst study to examine mutual monitoring. Their experiment used

18


19/24

dreoni (1988). Andreoni showed that contributions decayed as would be ex-pected by an equilibrium learning hypothesis, but contrary to learning also

showed that when the game was restarted contributions returned to signicantlevels. One explanation of this result is that reciprocating participants withholdcontributions to punish free riders, but are willing to wipe the slate clean whenthe experiment is restarted. More directly, Fehr & Gaechter show that whenparticipants have some way, other than withholding contributions, to punishfree riders, they do so and contributions increase.

The work of Fehr and Gaechter piqued the interest of other researchers whohave conrmed their main result and extended the analysis in other interest-ing directions. Bowles, Carpenter and Gintis (2001) develop a team productionmodel based on reciprocity which predicts punishment in equilibrium and testthe model experimentally. The experiments substantiate the major hypothesisgenerated by the model - transferring residual claimancy to a team increasesreciprocators propensity to punish shirkers and this, in turn, increases the pro-ductive efficiency of team production. Additionally, Carpenter (2001) showsthe effectiveness of mutual monitoring need not be attenuated in large groups.Page & Putterman (2000) also conrm that punishment is used to maintain orincrease contributions to a public good and show that communication amongplayers, which usually increases contributions, has mixed effects when combinedwith sanctions. Finally, Sefton, Shupp and Walker (2000) ran an experimentin which players could reward and sanction other players. When both rewardsand sanctions are allowed, they show that initially, rewards are used, but by theend of the experiment rewards abate and players rely mainly on sanctions.

The experiments discussed above demonstrate two behavioral facts: rst,subjects will punish others even at some cost to themselves. This is a veryrobust result seen in a variety of choice environments. Second, punishment is

used to elicit contributions in social dilemma situations. While these facts areimportant and relate to ingroup processes, the model we have in mind requirespunishment to be generalized when smaller groups are subsumed into largerpopulations. With this in mind, Carpenter, Matthews and Ongonga (2001)ask whether people internalize the heuristic of punishing free riders and simplypunish all deviations from the group norm, regardless of the group membershipof the norm violator. If we can demonstrate that punishment diffuses past smallgroup boundaries and can be maintained in a larger, more opaque population,then we have the behavioral foundations for generalized class action on a largescale.

Figure 4 presents data from the Carpenter, Matthews and Ongonga (2001)experiment. The top three lines plot the average level of contributions in threetreatments: the control which is a standard public goods game, the mutual

monitoring game in which participants can monitor and punish people in theirgroup, and the social reciprocity treatment where players can punish free ridersin a completely separate group, in addition to the members of their own group.

the common pool resource game in which players contribute by refraining from extracting acommonly held resource. This work has been extended in Ostrom, Walker and Gardner ( 1994)and Moir (1998).

19


20/24

1 2 3 4 5 6 7 8 9 10Period

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0Social Reciprocity (14 groups)

Mutual Monitor (6 groups)

Standard VCM (4 groups)

SR Ingroup Punishment

SR Outgroup Punishment

Mutual Monitor Punishment

Contributions and Punishment

FractionofEndowmentContributed&

FractionofEarningsSpentonPunishment

Figure 4: Average Contributions and Expenditures on Punishment (source:Carpenter & Matthews 2001)

The bottom three lines plot the average expenditure on punishment in the twopunishment treatments. Remarkably, not only do people punish free ridersin their own group which can be rationalized in terms of eye for an eye likereciprocity, they also punish free riders in the other group despite the fact thatthey can never benet from getting this person to contribute in the future.Furthermore, the free rider in the other group also imposed no harm (i.e. fromthe lower overall contribution level) on the punisher. The authors take thisa strong evidence in favor of the idea that, although punishment norms mayevolve in small ingroups, once internalized, punishing behavior can operate inlarger populations.

Also notice that punishment increases contributions in both treatments tolevels greater than the control which demonstrates the standard decline in con-tributions over trials. It is also the case that, social reciprocity increases con-tributions even more than simple ingroup punishment which implies that largepopulations in which people can punish not only their own group members butalso people transgressing a contribution norm in other groups achieve higher

payoff

s than populations in which punishment is localized.

20


21/24

5 Concluding Remarks

Contrary to the standard, prisoners dilemma representation of collective ac-tion attributed to Olson (1965) which explicitly challenges the theory of classconsciousness and class action offered by Marx, the Marxian approach seems tobe not too far from the mark when one employees more sophisticated modelingtools (e.g. bounded rationality and evolutionary game theory) and pays closeattention to the behavioral literature.

We have offered a view of economic and political agents who are purposivelike the standard representation, but allow them to learn and be subject to biasesthat have deep empirical support. When we combine our empirical regularities(e.g. an ingroup bias, a propensity to punishment etc.) with a model thatallows for imperfect agents we see that equilibria emerge in which collectiveaction occurs and is sustained.

While we feel we have made a strong case for the behavioral relevance of

Marxs concept of collective and class action, more work is planned. First, weanticipate our model can be improved by examining the link o ffered by Sabia(1988) more closely. Specically, we plan to model the interaction between theinternalization of norms in small groups and their propagation in larger popu-lations once internalized. To match our theory, we also plan to run experimentsthat would allow players to participate for some time in small groups and then,hopefully after norms evolve, form larger groups to see if small group processescan be adopted by larger groups.

6 References

Ahn, T.-K. (1999). "Interdependence and heterogeneity in the Finitely

Repeated PD Games." mimeo.Ameden, H., A. Gunnthorsdottir, et al. (1998). A Minimal Property Right

System for the Sustainable Provision of Public Goods mimeo. University ofArizona.

Anderson, S. P., J. K. Goeree and C. A. Holt (1998) A Theoretical Analy-sis of Altruism and Decision Error in Public Goods Games, Journal of PublicEconomics, Vol. 70, 297-323.

Andreoni, J. (1988). "Why Free Ride? Strategies and Learning in PublicGood Experiments." Journal of Public Economics 37: 291-304.

Axelrod, R. (1986) An Evolutionary Approach to Norms, American PoliticalScience Review, Vol. 80, 1095-1111.

Bendor, J. and D. Mookherjee (1987) Institutional Structure and the Logicof Ongoing Collective Action, American Political Science Review, Vol. 81, 129-154.

Binmore, K. et al (1999) A Backward Induction Experiment, UniversityCollege London Working Paper.

Blume, L. (1997) Population Games, in W. B. Arthur, S. N. Durlauf andD. Lane, eds, The Economy as an Evolving Complex System II. Redwood City:

21


22/24

Addison-Wesley.Bowles, S. and H. Gintis (1993) The Revenge of Homo Economicus: Con-

tested Exchange and the Revival of Political Economy, Journal of EconomicPerspectives, Vol. 7, 83-102.

Bowles, S. and H. Gintis (1998) How Communities Govern: The StructuralBasis of Prosocial Norms, in A. Ben-Ner and L. Putterman, eds, Economics,Values and Organization. New York: Cambridge University Press.

Bowles, S. and H. Gintis (1999) The Evolution of Strong Reciprocity, SantaFe Institute Working Paper

Bowles, S., J. Carpenter, et al. (2001). "Mutual Monitoring in Teams: TheEffects of Residual Claimancy and Reciprocity." mimeo.

Brewer, M. (1979). "In-Group Bias in the Minimal Intergroup Situation: Acognitive-motivational analysis." Psychological Bulletin 86: 307-324.

Brewer, M. and R. Kramer (1986). "Choice in Social Dilemmas: Effectsof social identity, group size and decision framing." Journal of Personality andSocial Psychology 50: 543-49.

Burawoy, M. (1989). Marxism Without Microfoundations, Socialist Review,Vol. 2, 53-85.

Carpenter, J. (1999). "Evolutionary Models of Bargaining: Comparingagent-based and discrete simulations of convention evolution." ComputationalEconomics forthcoming.

Carpenter, J. (2001). "Punishing Free-Riders: how group size affects mutualmonitoring and collective action." mimeo.

Carpenter, J., S. Burks, et al. (2001). "High Stakes Bargaining with Non-Students." mimeo.

Carpenter, J. and P. Matthews. (2001). "Social Reciprocity." mimeo.Chen, H. C., J. Friedman and J. F. Thisse, Boundedly Rational Nash Equi-

librium: A Probablistic Choice Approach, Games and Economic Behavior, Vol.18, 32-54.Chwe, M. S-Y. (1999) Structure and Strategy in Collective Action, American

Journal of Sociology, Vol. 105, 128-156.Coran, B. T. (1995) Some Applications of Game Theory to Marxian Political

Economy, Rethinking Marxism, Vol. 8, 48-62.Croson, R. (1996). "Partners and Strangers Revisited." Economics Letters

53: 25-32.Diermeier, D. and J. A. Van Mieghem (2000) Spontaneous Collective Action,

Northwestern University (MEDS) Working Paper.Eckel, C. and P. Grossman (2001). "Managing Diversity by Creating Team

Identity." mimeo.Elster, J. (1982) Marxism, Functionalism and Game Theory: The Case for

Methodological Individualism, Theory and Society, Vol. 11, 453-482.Fehr, E. and S. Gaechter (2000). "Cooperation and Punishment in Public

Goods Experiments." American Economic Review 90(4): 980-994.Foster, J. (1973) Class Struggle and the Industrial Revolution. New York;

Metheun.

22


23/24

Friedman, J. (1971) A Non-Cooperative Equilibrium for Supergames, Reviewof Economic Studies, Vol. 38, 1-12.

Fudenberg, D., D. Levine and E. Maskin (1994) The Folk Theorem withImperfect Public Monitoring, Econometrica, Vol. 62, 997-1039.

Granovetter, M. (1978) Threshold Models of Collective Behavior, AmericanJournal of Sociology, Vol. 83, 1421-1442.

Griesinger, D. and J. Livingston (1973). "Toward a Model of InterpersonalMotivation in Experimental Games." Behavioral Science 18: 173-188.

Guttman, J. M. (2000) On the Evolutionary Stability of Preferences forReciprocity, European Journal of Political Economy, Vol. 16, 31-50.

Keser, C. and F. van Winden (2000). "Conditional Cooperation and Vol-untary Contributions to Public Goods." Scandinavian Journal of Economics102(1): 23-29.

Kirman, A. P. (1997) The Economy as an Interactive System, in W. B.Arthur, S. N. Durlauf and D. Lane, eds, The Economy as an Evolving ComplexSystem II. Redwood City: Addison-Wesley.

Kollock, P. (1998). Transforming Social Dilemmas: Group identity andcooperation. Modeling Rational Moral Agents. P. Danielson. Oxford, OxfordUniversity Press.

Komorita, S. and C. Lapworth (1982). "Cooperative Choice Among Individ-uals versus Groups in an N-person Dilemma Situation." Journal of Personalityand Social Psychology 42: 487-96.

Kramer, R. and M. Brewer (1984). "Effects of Group Identity on Resourceuse in a Simulated Commons Dilemma." Journal of Personality and Social Psy-chology 46(5): 1044-57.

Ledyard, J. O. (1995) Public Goods: A Survey of Experimental Research,in J. H. Kagel and A. Roth, eds, The Handbook of Experimental Economics.

Princeton: Princeton University Press.Mansbridge, J. (2000) Oppositional Consciousness, Part II. in J. Mansbridgeand A. Morris, eds, Oppositional Consciousness. Chicago: University of ChicagoPress (forthcoming).

Marx, K. (18xx) The Eighteenth Brumaire of Louis Napoleon.Marx, K. and F. Engels (18xx). The Communist Manifesto.Matthews, P. H. (2000a) Did Marx Know the Way to Santa Fe? in D.

Colander, editor, Complexity and the History of Economic Thought. New York:Routledge, 2000.

Matthews, P. H. (2000b) Technical Change and the Evolution of ClassConscious Norms, Review of Radical Political Economics, Vol. 32, 470-481.

McBride, M. (2001) Discrete Public Goods under Threshold Uncertainty,Yale University mimeo.

McClintock, C. and W. Liebrand (1988). "Role of Interdependence Struc-ture, Individual Value Orientation, and Anothers Strategy in Social DecisionMaking: a transformational analysis." Journal of Personality and Social Psy-chology 55(3): 396-409.

McKelvey, R. D. and T. R. Palfrey (1995) Quantal Response Equilibria forNormal Form Games, Games and Economic Behavior, Vol. 10, 6-38.

23


24/24

Michl, T. and J. Baldani (2000) Technical Change and Prots: The Pris-oners Dilemma, Review of Radical Political Economics, Vol. 32, 104-118.

Moir, R. (1998). "Spies and Swords: Costly Monitoring and Sanctioning ina Common-Pool Resource Environment." mimeo.

Muntaner, C. (1994) Class Behavior: An Alternative to Rational ChoiceMicrofoundations, Rethinking Marxism, Vol. 7, 118-131.

OConner, J. (1973) The Fiscal Crisis of the State. New York: St. Martins.Oliver, P. E. (1993) Formal Models of Collective Action, Annual Review of

Sociology, Vol. 19, 271-300.Olson, M. (1965) The Logic of Collective Action. Cambridge: Harvard Uni-

versity Press.Ostrom, E., R. Gardner, et al. (1994). Rules, Games and Common-Pool

Resources. Ann Arbor, University of Michigan Press.Ostrom, E., J. Walker, et al. (1992). "Covenants With and Without a

Sword: self-governance is possible." American Political Science Review 86: 404-17.

Page, T. and L. Putterman (2000). "Cheap Talk and Punishment in Volun-tary Contribution Experiments." mimeo.

Posner, R. (1977) Economic Analysis of the Law. Boston: Little Brown.Resnick, S. and R. D. Wolff (1982). Classes in Marxian Theory, Review of

Radical Political Economics, Vol. 13, 1-18.Rosner, P. (1998) Karl Marx: A German Economist, History of Economics

Society (HES) Conference Paper.Runge, C. F. (1984) Institutions and the Free Rider: The Assurance Problem

in Collective Action, Journal of Politics, Vol. 46, 154-181.Sabia, D. R. (1988) Rationality, Collective Action and Karl Marx, American

Journal of Political Science, Vol. 32, 50-71.

Sefton, M., R. Shupp, et al. (2000). "The Eff

ect of Rewards and Sanctionsin Provision of Public Goods." mimeo.Sen, A. (1973) On Economic Inequality. Oxford: Clarendon Press.Sethi, R. (1996) Evolutionary Stability and Social Norms, Journal of Eco-

nomic Behavior and Organization, Vol. 29, 113-140.Sethi, R. and E. Somanathan (2000) Understanding Reciprocity, Barnard

College Working Paper.Skyrms, B. (1996). Evolution of the Social Contract. New York, Cambridge

University Press.Taylor, M. (1976). Anarchy and Cooperation. London, John Wiley & Sons.Taylor, P. and L. Jonker (1978) Evolutionarily Stable Strategies and Game

Dynamics, Mathematical Biosciences, Vol. 40, 145-156.

24

behavioral marxism i: collective action

Documents